Research map / Detection and segmentation / Open-set and few-shot detection
Learning to Diversify and Focus: A Reinforcement Framework for Open-Vocabulary HOI Detection
Yongchao Xu, Jiawei Liu, Junfeng Wang, Sen Tao, Na Jiang, Zheng-Jun Zha
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add Semantic Diversification (SD) module (ablation) | CLIP-based OV-HOI baseline · SWIG-HOI (Unseen) | mAP: 7.35 -> 14.01 | Improved |
| add Interaction Focusing (IF) module (ablation) | CLIP-based OV-HOI baseline · SWIG-HOI (Full) | mAP: 11.47 -> 17.64 | Improved |
| add IF on top of SD (ablation) | SD-IF · SWIG-HOI (Full) | mAP: 16.89 -> 19.48 | Improved |
| add IF on top of SD for unseen classes (ablation) | SD-IF · SWIG-HOI (Unseen) | mAP: 14.01 -> 15.57 (IF alone 13.50 vs SD alone 14.01) | Mixed |
| add Global Expansion to SD (with Local Consistency) (ablation) | SD module · SWIG-HOI (Unseen) | mAP: 13.79 -> 15.57 | Improved |
| add Local Consistency to SD (with Global Expansion) (ablation) | SD module · SWIG-HOI (Unseen) | mAP: 15.43 -> 15.57 | Improved |
| add Spatial Focusing Reward to IF (with Semantic Consistency Reward) (ablation) | IF module · SWIG-HOI (Full) | mAP: 17.67 -> 19.48 | Improved |
| add Semantic Consistency Reward to IF (with Spatial Focusing Reward) (ablation) | IF module · SWIG-HOI (Full) | mAP: 19.32 -> 19.48 | Improved |
| use RL policy instead of Gaussian noise injection in SD (ablation) | SD module · SWIG-HOI (Unseen) | mAP: 9.86 -> 14.01 | Improved |
| use RL actor-critic instead of supervised attention regression in IF (ablation) | IF module · SWIG-HOI (Unseen) | mAP: 12.65 -> 13.50 | Improved |
| use the paper's lightweight actor-critic instead of PPO (ablation) | IF module · SWIG-HOI (Full) | mAP: 18.32 -> 19.48 | Improved |
| use the paper's lightweight actor-critic instead of SAC (ablation) | IF module · SWIG-HOI (Full) | mAP: 18.48 -> 19.48 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SGC-Net — SGC-Net: Stratified Granular Comparison Network for Open-Vocabulary HOI Detection (CVPR 2025) · better on 7 of 7 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Open-Vocabulary HOI Detection with Interaction-aware Prompt and Concept Calibration (ICCV 2025 · similarity 0.92)
- Exploring the Potential of Large Foundation Models for Open-Vocabulary HOI Detection (CVPR 2024 · similarity 0.91)
- Open-World Human-Object Interaction Detection via Multi-modal Prompts (CVPR 2024 · similarity 0.91)
- SGC-Net: Stratified Granular Comparison Network for Open-Vocabulary HOI Detection (CVPR 2025 · similarity 0.90)
- Visual Diversity and Region-aware Prompt Learning for Zero-shot HOI Detection (NeurIPS 2025 · similarity 0.90)
- HOIMamba: Efficient Mamba-based Disentangled Progressive Learning for HOI Detection (AAAI 2025 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
